Tools
Tools
Haouam, Kamel Eddine (2008) RVSM: a rhetorical conceptual model for content-based indexing and retrieval of text document. Doctoral thesis, London Metropolitan University.
The amount of information available on the Internet is currently growing at an incredible rate. However, the lack of efficient indexing is still a major barrier to effective and accurate information retrieval on the web. This dissertation introduces a semantic model for content-based indexing and retrieval of relevant text documents from a large collection such as the Internet. It presents the investigation and adaptation for a conceptual technique for indexing and retrieval of relevant documents by content from large collection of text documents. The technique aims at improving the quality of retrieval by capturing the semantics of the documents. It performs the conventional keyword based indexing and introduces a thematic relationship between parts of text adapting a linguistics theory called Rhetorical Structure Theory (RST). RST is mostly used for summarizing and text generation but in this work it is adapted to the indexing. The technique enhances conventional keyword indexing with the linguistic theory called RST, which is based on the identification of relations amongst parts of the text using cue phrases or words (however, thus, so, but, etc... ). All other techniques of Information Retrieval (IR.) through away these cue words but we use them to discover semantics for indexing. In this thesis, an Index/Retrieval model called "Rhetorical Vector Space Model" (RVSM) was developed by combining traditional statistical/probabilistic methods (Salton' s Vector Space Model (VSM)) with semantic methods. In this model, both documents and queries are represented as rhetorical matrices. A search mechanism was designed to compute the similarity between two rhetorical matrices and the similarity value was interpreted as the predictor of relevancy. The conceptual model was described and a prototype concept was built to prove the validation of the Index/Retrieval model by comparing against existing indexing systems. The prototype system was used in a series of experiments designed to evaluate the proposed technique in terms of precision, recall, and effectiveness of relevance ranking. The results of these experiments showed that the system based on our technique had significantly better performance than the other existing techniques. It will perform better than the statistical methods alone.
 |
View Item |